What you can - and can't - hope to understand about how AI works

This page has 0 threads | Add post

Blogs

Finding our blogs useful? Please consider helping us to share more!

What you can - and can't - hope to understand about how AI works Part one of a four-part series of blogs
I'm sick of reading blogs pretending to explain how Large Language Models work, when the truth is that few people on earth can ever hope to understand how they are constructed. This blog is my attempt to explain as much as 99.99% of us can ever hope to understand about how AI works, and why the other bits will always remain opaque to most people. What you can - and can't - hope to understand about how AI works (this blog) Tokenisaton and embedding Large language models Constraints on the use of AI

What you can - and can't - hope to understand about how AI works

Part one of a four-part series of blogs

I'm sick of reading blogs pretending to explain how Large Language Models work, when the truth is that few people on earth can ever hope to understand how they are constructed. This blog is my attempt to explain as much as 99.99% of us can ever hope to understand about how AI works, and why the other bits will always remain opaque to most people.

What you can - and can't - hope to understand about how AI works (this blog)
Tokenisaton and embedding
Large language models
Constraints on the use of AI

Posted by Andy Brown on 16 July 2024

The aim of this blog is:

To explain what's explainable about how AI works.
To show why you can never hope to understand the other bits.
Given the above, to show the constraints you'll come up against when using AI tools.

Just in case you're feeling patronised or belittled, a bit of context. I have a strong background in maths, computing and probability, but I know that I can never hope to understand how AI tools work. I just want to speed your journey to the same conclusion!

When you submit a question to an AI tool, here's what happens behind the scenes:

A simple diagram showing what happens when you submit a question to an AI tool like ChatGPT, Claude, Copilot or Gemini.

In the rest of this blog, i'll look at the middle parts of this process, and then explain why the nature of what's going on implies some constraints on how you can use AI tools in your day-to-day work. Here's what you'll learn in this blog:

Part of blog	Notes
Tokenisation and embedding	An explanation of how AI tools turn your questions into numbers that tthe AI tools can understand.
Large language models (LLMs)	How large language models are built and used, and why you can never hope to understand much about them.
Constraints on AI	Given what you've learned about how AI tools work, the main constraints you will come up against.

Let's start with the "easy" bit - tokenisation.

Parts of this blog
What you can - and can't - hope to understand about how AI works (this blog) Tokenisaton and embedding Large language models Constraints on the use of AI